Sudhanshu Verma

How much longer will we need to design interfaces?

22 September 2026

If an agent can use software for me, how much of its interface will I ever see?

I keep coming back to this question. A lot of interface design exists because people need to operate software. We organise menus, make features discoverable and shorten the path from wanting something to getting it done.

But what happens when people stop taking that path themselves?

Think about Blender or After Effects. Getting good at these tools can take years. You’re learning the craft, but also how to translate an idea into the software’s language: layers, nodes, keyframes, panels. If I can ask an agent to light a scene or animate a headline, then refine the result with it, I may never need to learn much of that machinery. And the less I need to operate it, the less its interface matters to me.

It’s a little like being a passenger. The dashboard still matters to whoever is driving, but it’s no longer how you get where you want to go. You give a destination.

An agent might initially operate the same interface we do. Over time, though, software could expose its capabilities directly to agents. The menus wouldn’t need to be part of that exchange at all.

Meanwhile, the interfaces we still need could become much easier to make well. Typography, spacing and interaction patterns are already partly packaged into design systems. AI could make more of that craft available without someone having to learn it from scratch.

That combination makes me question how much interface design work will remain: fewer screens to operate, and less specialist effort to create them.

Some interfaces will earn their place. Comparing images side by side or dragging a keyframe can be easier than explaining what you want. Sometimes using the tool is how you discover the idea. I expect us to keep those interactions.

But their survival doesn’t guarantee the relevance of everything around them, or enough work to sustain the profession as it exists today.

As a designer, I think we need to take that possibility seriously. Before asking how to make an interface better, we may increasingly have to explain why someone needs to use it at all.